Schizophrenia is a severe psychiatric disorder that is associated with staggeringly high individual and societal costs. Although the illness is typically chronic, it is not static, and the majority of people with schizophrenia vacillate between full or partial remission and episodes of symptomatic relapse. Relapses increase one's risk for major problems including homelessness, incarceration, victimization, and suicide. Moreover, patients with schizophrenia who relapse are three to four times more costly than those who do not. The goal of the proposed project is to develop and evaluate a novel paradigm for illness monitoring, detection of early warning signs, and relapse prevention in schizophrenia. Our interdisciplinary team of clinical researchers and computer scientists proposes to develop a mobile system that uses smartphone-embedded sensors (i.e. microphone, accelerometer, GPS, light sensor) coupled with computerized self-reports, to track a range of behaviors (i.e. paralinguistic aspect of speech, physical activity, location, sleep, mood, psychotic symptoms) that are relevant to relapse in schizophrenia. Using machine learning techniques, the system will leverage behavioral data and patient self-reported clinical updates to generate personalized early warning models. The models will evolve with use of the system over time, focusing on variability from one's typical behavioral patterns to calibrate a unique patient relapse signature. Treatment teams will be informed about patients' clinical status via secure website. When the mobile system flags trends that are consistent with one's relapse signature, it will trigger patient functions and provider functions (i.e. real-time notification, prompts to initiate contact, time-sensitive treatments) to help prevent progression to full psychotic relapse. In Phase 1 of the project, we will integrate multi-modal sensor, ecological momentary assessment, and machine learning technologies into a unified smartphone system that will be tested and refined in laboratory settings. In Phase 2, we will conduct field trials with individuals with schizophrenia i real-world conditions to identify and resolve technical and mechanical problems, adapt the software, and maximize system usability. In Phase 3, we will conduct a randomized 12- month trial of the monitoring and prevention system compared to treatment as usual in 150 outpatients with schizophrenia that are at high-risk for relapse. If successful, our proposed system can be rapidly made available to a population that is in dire need of more effective resources, and can serve as a template for mobile monitoring and treatment systems for a range of clinical conditions with an episodic nature.